Survival analysis to measure turnover of the medical education workforce in Ethiopia.

Background: Until recently, there were only a few medical schools in Ethiopia. However, currently, in response to the apparent shortage in physician workforce, the country has made huge progress with respect to the expansion of medical schools, by adopting the so-called flooding strategy. Nevertheless, the effectiveness of the intended strategy also relies on physician accessibility and turnover. Therefore, the aim of this study was to examine the distribution of physicians in the medical schools of Ethiopia and to quantify the magnitude and identify factors associated with physician turnover.

Methods: This organizational faculty physician workforce survey was conducted in seven government-owned medical schools in Ethiopia. Longitudinal medical workforce data set of about 6 years (between September 2009 and June 2015) were retrospectively collected from each of the medical schools. The observation time begins with the date of employment (time zero) and ends at the date on which the physician leaves the appointment/or the data collection date. Kaplan-Meier survival method was used to describe the duration of stay of physicians in the academic health care settings. A Cox proportional hazards (CPH) model was fitted to identify the risk factors for physician turnover.

Results: In this study, a total of 1258 faculty physicians were observed in seven medical schools which resulted in 6670.5 physician-years. Of the total, there were 198 (15.7%) turnover events and the remaining 1060 (84.3%) were censored. The average turnover rate is about 29.7 per 1000 physician-years of observations. Multivariate modeling revealed no statistical significant difference in the rate of turnover between males and females (adjusted hazard ratio (AHR), 1.12; 95%CI, 0.71, 1.80). However, a lower rate of physician turnover was observed among those who were born before 1975 (AHR, 0.37; 95%CI, 0.20, 0.69) compared with those who were born after 1985. Physicians with the academic rank of associate professor and above had a lower (AHR, 0.25; 95%CI, 0.11, 0.60) rate of turnover in comparison to lecturers. In addition, physicians working in Jimma University had 1.66 times higher rate of turnover compared with those working in Addis Ababa University. However, the model showed a significantly lower rate of turnover in Mekelle (AHR, 0.16; 95%CI, 0.06, 0.41) and University of Gondar (AHR, 0.46; 95%CI, 0.25, 0.84) compared with that of Addis Ababa. Physician turnover in the remaining medical schools (Bahir Dar, Haromaya, and Hawassa) did not show a statistically significant difference with Addis Ababa University (P > 0.05).

Conclusions: This study revealed a strong association between physician turnover with age, academic rank, and workplace. Therefore, the findings of the study have important implications in that attention needs to be given for the needs of faculty physicians and for improving the work environment in order to achieve a high level of retention.